Low-rank and joint-sparse signal recovery using sparse Bayesian learning in a WBAN

Wireless body area networks (WBANs) will become increasingly important in future communication systems, especially in the area of wearable health monitoring systems, such as telemonitoring systems for the collection of electrocardiogram (ECG) data/electroencephalogram (EEG) data via WBANs for e-health applications. However, wearable devices usually require limited power consumption to ensure long battery life. Fortunately, compressed sensing (CS) has been proven to use less energy than traditional transform-coding-based methods. Because the spatial and temporal data collected by a WBAN have some closely correlated structures in certain transform domains (e.g., the discrete cosine transform (DCT) domain), we exploit these structures to propose a new low-rank and joint-sparse (L&S) signal recovery algorithm for recovering ECG/EEG data in the framework of CS. Using a simultaneously L&S signal model, we employ a Bayesian learning treatment. This treatment incorporates an L&S-inducing prior over the data and appropriate hyperpriors over all hyperparameters and thereby yields an effective reconstruction of L&S data. Simulation results with synthetic and real ECG/EEG data demonstrate that the proposed algorithm is superior to other state-of-the-art recovery algorithms in terms of reconstruction performance with comparable computational complexity.